Page MenuHomePhabricator

ircecho does not handle netsplits well
Closed, ResolvedPublic

Description

gerrit-wm appears to no longer be reporting changes to Gerrit in MediaWiki-General.


Version: wmf-deployment
Severity: major

Details

Reference
bz43112

Event Timeline

bzimport raised the priority of this task from to High.Nov 22 2014, 12:46 AM
bzimport set Reference to bz43112.

reloaded it on manganese and now it's fine. perhaps it didn't do well with the netsplit yesterday.

Re-opening as this appears to be happening again.


gerrit-wm!~gerrit-wm@manganese.wikimedia.org
ircname IRC echo bot
channels : #wikimedia-dev MediaWiki-General #mediawiki-parsoid #wikimedia-mobile +#wikimedia-operations
server holmes.freenode.net London, UK

idle : 1 days 1 hours 51 mins 54 secs signon: Mon Dec 31 17:47:25 2012

restarted the bot, I wont' know if it's happy til there's a change though.

The bot really ought not suddenly go quiet. It's staying connected to the network and joined to the channels, it just (seemingly) stops sending messages to the channel inexplicably. Re-opening this bug and updating its summary accordingly.

(In reply to comment #5)

The bot really ought not suddenly go quiet. It's staying connected to the
network and joined to the channels, it just (seemingly) stops sending
messages
to the channel inexplicably.

I don't usually idle on MediaWiki-General so I can't speak of other times, this one it was because of a netsplit at 2012-12-31 22.46 UTC from which it recovered a minute later: last message is a minute before it and afterwards silence.
wikibugs is less capriccioso.

(In reply to comment #6)

I don't usually idle on MediaWiki-General so I can't speak of other times, this one
it was because of a netsplit at 2012-12-31 22.46 UTC from which it recovered a
minute later: last message is a minute before it and afterwards silence.
wikibugs is less capriccioso.

Ariel explained to me that there's an outstanding bug in irclib (or ircbot.py or whatever gerrit-wm uses) in handling netsplits. I suppose this bug is as good as any to track that issue.

(In reply to comment #7)

Ariel explained to me that there's an outstanding bug in irclib (or ircbot.py
or whatever gerrit-wm uses) in handling netsplits. I suppose this bug is as
good as any to track that issue.

I suspected it was a problem with the bot and not actually Gerrit. I'm going to move it to General since it's not a Gerrit problem (and presumably affects other services using ircecho).

Already deployed for gerrit, icinga, and icinga (labs). I'm marking this as fixed for now. If the bot still has problems we can reopen.